llm-wiki

Build knowledge that compounds — not answers that evaporate.

A git-backed wiki engine that turns a folder of Markdown files into a searchable, structured knowledge base. Accessible from the command line, from any MCP-compatible agent, or from any IDE via ACP.

The engine has no LLM dependency. It manages files, git history, full-text search, and knowledge structure. The LLM is always external.

Install

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/geronimo-iia/llm-wiki/main/install.sh | bash

# Windows (PowerShell)
irm https://raw.githubusercontent.com/geronimo-iia/llm-wiki/main/install.ps1 | iex

# Or via cargo
cargo install llm-wiki-engine

→ All installation options

Quick Start

# Create a wiki
llm-wiki spaces create ~/wikis/research --name research

# Start the MCP server
llm-wiki serve

Connect your editor (VS Code, Cursor, Windsurf, Zed, Claude Code), then use the tools to create pages, ingest sources, search, and build knowledge.

→ Getting started guide

Why Not RAG?

Most AI knowledge tools retrieve and generate on every query. Knowledge doesn't accumulate.

llm-wiki implements a Dynamic Knowledge Repository (DKR): process sources at ingest time, not query time. The LLM integrates each source into the wiki — updating concept pages, creating source summaries, flagging contradictions — and commits the result. Knowledge compounds with every addition.

What It Does

• Search — full-text BM25 search across one or all wikis
• Type system — JSON Schema validation per page type, field aliasing, custom types via schemas/
• Concept graph — typed nodes, labeled edges (fed-by, depends-on, cites), Mermaid and DOT output
• Git-backed — every change is a commit, full history, diffable
• Multi-wiki — manage multiple wikis from one process
• MCP + ACP — stdio, HTTP, and ACP transports for any agent

What It Is Not

• Not an LLM — makes no AI calls
• Not a RAG system — does not retrieve and generate on demand
• Not a note-taking app — it is an engine, you bring your own interface
• Not a static site generator — but llm-wiki-hugo-cms can render the wiki as a website

Technology

Single Rust binary. No runtime, no database, no Docker.

Documentation

Related Projects

Acknowledgments

• Andrej Karpathy — for the LLM Wiki gist that defined the Dynamic Knowledge Repository pattern.
• vanillaflava — for llm-wiki-claude-skills, which turned the pattern into a practical skill architecture.

llm-wiki is a continuation of agent-foundation.

Contributing

Contributing guide · Code of conduct · Security policy

License

MIT OR Apache-2.0